Satya Nadella has a message for every company reaching for the shiniest AI tool in the drawer: stop it.
The Microsoft CEO, speaking on the New York Times Hard Fork podcast in June 2026, urged businesses to rethink their reflexive dependence on the most powerful AI models for tasks that don’t require them. His core argument is deceptively simple. If AI’s economic benefits don’t flow broadly across industries, the entire sector risks looking like a bubble, and political backlash won’t be far behind.
The case against token-maxing
Nadella coined a useful term for the problem he’s diagnosing: “token-maxing.” It’s the tendency among companies to throw the most advanced, most expensive AI models at every problem, regardless of whether a simpler, cheaper model would do the job just fine.
“Don’t use frontier models for non-frontier problems,” Nadella said during the podcast appearance.
